Stata - How to conduct OLS regressions

[@damiainwhite4984]

How would you answer a question, which asked to find the relationship between two variables, and how that relationship depends on another independent variable? For example, what effect does the proportion of female workers have on firm performance, depending on education levels of the female?

[@SteffensClassroom]

Dear Damiain Thank you for your question! The easiest answer I can get you is to watch any lecture 1 of my research methods in Finance course on this channel. There I address 'this' question. Have a great day!

[@damiainwhite4984]

@@SteffensClassroom Perfect, will have a look at the lecture. Thanks for your help.

[@SteffensClassroom]

You're welcome!

[@thehappyloaf]

Is it possible to perform stepwise regression on stata?

[@SteffensClassroom]

Possible for sure. However, I didn't cover it yet :)

[@thehappyloaf]

@@SteffensClassroom is there another morr robust way to perform the regression? I have a hundred independents and I'm sure there will be multicollinearity so how do I ensure the model drops the variables it needs to? Without iterating myself

[@SteffensClassroom]

​@@thehappyloafStat will automatically drop variables that are perfectly colinear. However, if you have that many variables, you want to look into ways of reducing this number in order to save degrees of freedom, and overall simplify your model. Factor analysis, principle component analysis or another summary index technique.

[@thehappyloaf]

@@SteffensClassroom Thank you! I will look into these techniques

[@FemkeHuisman]

If you add a dummy variable to a regression (not in the way you showed with foreign==1), should you always put i. in front of it?

[@SteffensClassroom]

In short: You don't have to. You can obtain the same via generate as with i. www.stata.com/support/faqs/data-management/creating-dummy-variables/ Related to this: Stata handles factor variables quite nicely, as you can read more about here: www.stata.com/features/overview/factor-variables/ I should make a video soon dedicated just to dummy variables and factor variables.

[@FemkeHuisman]

@@SteffensClassroom Thank you. I have one more question, regarding descriptive statistics. If I make a table, I get different "N's" (number of observations) for different variables. Does this matter, and if yes, how can I get the sample to only show the lowest N so that only actual observations are included? I do not want to drop things I shouldn't....

[@SteffensClassroom]

Say you run a regression, then Stata uses list-wise deletion by default. That is, if an observation in your dataset has a missing value for one or more of your variables, then it is not 'included' in your regression. Somebody can correct me if I am wrong, but don't delete anything if you don't have to, and if you do, clearly document it. Always be as transparent as possible!

[@FemkeHuisman]

@@SteffensClassroom Yes, for the regression that is true. But now I'm only making a descriptive statistics table! So there, for example, one variable has n = 773, whereas another has n=835. I would like them to all have the same (so, lowest) number!

[@FemkeHuisman]

Or does it not matter for a descriptive statistics table?